JP4548218B2 - Face recognition device - Google Patents

Description

  The present invention relates to a face authentication apparatus that performs authentication by performing template matching between a face image of a person registered in advance as a template image and a face photo of a person photographed for authentication.

  Conventionally provided general face authentication devices, for example, register a face image of a person who is permitted to enter the room as a template image in advance, and then capture the face of the person who wants to enter the room with an imaging camera. Thus, the authentication is performed by determining whether the photographed face image is the same as a pre-registered template image using template matching by image processing.

  However, in the conventional face authentication device, the face image input from the imaging camera as a face image of the person who is about to enter the room is a photograph of a human face of a living body, or a face photograph held in front of the imaging camera It is not judged whether the image was taken by the imaging camera.

  Therefore, the face authentication apparatus is such that a person who is not permitted to enter the room is in front of the face authentication apparatus using the face photograph of the person who is permitted to enter the room. If it is assumed that the person is admitted, even if it is determined that the person is allowed to enter the room even if it is a facial photograph, it will not function as a security device. The same applies to faces and mannequins as well as face photos.

Therefore, in order to solve the above-described problem, a personal recognition device that detects a movement of a facial part and recognizes an individual is provided (Patent Document 1) .
The method adopted in this personal recognition device blinks eyes if it is a living body face, but does not blink eyes if it is a face of a photograph. This is a method for judging whether or not there is a change in movement.

There is also provided a face image processing system that employs a method for discriminating whether the subject is a living body or a facial photograph by determining whether or not the living body is based on a temporal change in pixel value by paying attention to a change in the pixel value of the eye part. (For example, Patent Document 2)
Japanese Patent Laid-Open No. 11-339048 (paragraph numbers 0041 to 0042) JP 2001-126091 (paragraph numbers 0078 to 0079)

  According to the device disclosed in Patent Document 1 described above, a change in the eye position is detected from the inter-frame difference image when it moves slightly, whether it is a living body face or a facial photograph. Therefore, even if it is a face photograph, it is determined that the person is a living person, and there is a possibility that the determination is erroneous.

  Further, even in the method disclosed in Patent Document 2, if the entire photograph moves, the pixel value changes and appears, so that there is a possibility of misjudging as in the method disclosed in Patent Document 1.

  Further, in the apparatus disclosed in Patent Document 1 described above, all processes are stopped when it is determined that the photograph is not a living person, and the face photograph is used by another person without being present. I couldn't determine who was impersonated, that is, who was photographed. Many of the faces shown in the photographs used in this way are faces of someone who is allowed to enter the room, and there are cases where a malicious person impersonates the person and tries to enter the room.

  The present invention has been made in view of the above points, and the object of the present invention is to determine whether the face of a person to be authenticated is a biological face or a pseudo face such as a photograph. It is an object of the present invention to provide a face authentication device capable of confirming a person who is a subject of impersonation later while preventing the above.

In order to achieve the above object, according to the first aspect of the present invention, there is provided a photographing camera for continuously photographing a plurality of frames including the face of the person to be authenticated for at least one frame, and an image of at least one frame from the imaging camera. Authentication means for determining whether or not a person to be authenticated is registered by template matching with a person image registered in advance as a template image, and a plurality of frames taken by the photographing camera on the face. A biometric identification unit that detects whether or not a subject of the photographing camera is a living body by detecting the presence or absence of movement of a plurality of parts; an authentication result indicating that the authentication unit is a registered person; and the biometric identification unit is a living body And a determination result output means for outputting a determination result that the person to be authenticated is a registered person only when the identification result is output, and Regardless of the discrimination result of the body identification means comprises a holding means for said recognizing means stores information for identifying the registered person when the person to be authenticated is authenticated as the registered person, the biometric identification The means extracts a face position from each image of a plurality of frames taken by the imaging camera, counts the number of change pixels in the eye area and the mouth area from the inter-frame difference image of the image at the face position, and the count number It is determined that there is movement when it exceeds a preset threshold corresponding to each area, and there is no movement when it does not exceed it, and when it is determined that only one of both areas is moving, An identification result indicating that it is present is output .

  According to the first aspect of the present invention, it is identified whether or not the face to be authenticated is a living body, and even if the authentication result of the authentication means is a registered person, authentication for impersonation using a face photograph or the like showing the face By confirming the information that identifies the registered person who is the target of impersonation saved by the holding means, the person can be alerted, and identification / authentication can be tightened against re-impersonation It becomes possible to cope with such.

Moreover , since the living body is identified by capturing the movement of the mouth and eyes that move together when the facial photograph is moved, the living body can be reliably identified.

According to a second aspect of the invention, in the first aspect of the invention, the biometric identification unit has a straight line detection function for detecting a straight line from an inter-frame difference image created from a plurality of frame images photographed by the photographing camera, When the straight line detection function detects a straight line having a predetermined length or longer, an identification result indicating that it is not a living body is output.

In the invention of claim 2 , impersonation can be prevented by identifying a facial photograph.

According to a third aspect of the present invention, in the first or second aspect of the invention, the biometric identification unit extracts a face position from each image of a plurality of frames taken by the imaging camera and creates an inter-frame difference image. If there are a predetermined number or more of change pixels in the vicinity of the outside of the face position in the inter-difference image, an identification result indicating that it is not a living body is output.

According to the third aspect of the present invention, it is possible to reliably identify a case where a face photograph is held in front of the face and impersonation is attempted.

According to a fourth aspect of the present invention, in any one of the first to third aspects, the authentication threshold setting means for lowering the authentication threshold of the authentication means to perform authentication when the identification result of the biometric identification means is not a living body. It is characterized by having.

According to the invention of claim 4 , it becomes easy to specify a person from a face photograph.

According to a fifth aspect of the present invention, in the invention according to any one of the first to fourth aspects, when the identification result of the biometric identification unit is not a biometric, a person who has authenticated that the authentication unit is a registered person is An identification threshold setting unit is provided that increases the identification threshold of the biometric identification unit for the person when the person is to be authenticated.

According to the invention of claim 5 , even if a malicious person attempts to impersonate again using the same face photo or the like, it is possible to prevent erroneous authentication by this impersonation and maintain security.

In the invention of claim 6, in the invention of any one of claims 1 to 5 , if the identification result of the biometric identification means is not a living body, the person who has authenticated that the authentication means is a registered person is An authentication threshold setting unit is provided for performing authentication by raising the face authentication threshold of the authentication unit when the person becomes an authentication target.

According to the invention of claim 6 , even if a malicious person attempts to impersonate again using the same facial photograph or the like, erroneous authentication by this impersonation can be prevented and security can be maintained.

The present invention identifies whether a face to be authenticated is a living body, and can prevent authentication against impersonation using a face photograph or the like that represents a face even if the authentication result of the authentication means is a registered person. In addition, by confirming the information that identifies the registered person who is the target of impersonation saved by the holding means, it is possible to take measures such as calling attention to the person or making identification / authentication stricter against impersonation again It has the effect of becoming. Moreover, since the living body is identified by capturing the movement of the mouth and eyes that move together when the facial photograph is moved, the living body can be reliably identified.

Embodiments of the present invention will be described below.
(Embodiment 1)
As shown in FIG. 1, the face authentication apparatus according to the present embodiment registers and holds face images of persons who are allowed to enter a room (building), for example, as template images in advance and register each personal information. A / D conversion is performed on the holding means 1, the imaging camera 2 that captures a plurality of frames of the face of the person H to be authenticated in time series, and the image signal of the upper body including the face captured by the imaging camera 2. A / D conversion means 3 for converting to a digital face image signal, a face image is extracted from an image of at least one frame of the A / D converted image, template matching is performed on the template image, and the extracted face Using the authentication means 4 for determining whether the person in the image is the same as any person whose face image is registered as a template image, and using the A / D converted face images of a plurality of frames The biometric identification means 5 for determining whether or not the face shown in the face image is a living body, the authentication means 4 outputs the result of authentication that it is a registered person, and the biometric identification means 5 is identified as a living body. Only when the result is output, the determination result for permitting the person H to enter the room is output. In other cases, the determination result for not permitting the person to enter the room is output, and the authentication is performed regardless of the identification of the living body. The judgment result output means 7 for holding the registered person's personal information, biometric identification result and input image in the holding means 6, and the voice recorded in advance when the photographing camera 2 starts photographing the face image. The sound generation means 8 that requests the subject to be blinked or open / close the mouth, and the holding means 6 are configured. Note that the instruction manual of the device has a description that “there is a biometric identification function, so that at the time of face authentication, blinking or opening / closing of the mouth is necessary”. It is unnecessary.

  Next, the operation of this embodiment will be described.

  First, when an appropriate person detecting means detects that a person stands in front of the photographing camera 1 or when an operation such as a photographing start button is performed, the photographing camera 1 performs an operation of photographing a plurality of images of the upper body of the subject. Do. At this time, the voice generation means 8 emits a voice requesting blinking or opening / closing of the mouth, thereby prompting the person to be photographed to blink or open / close the mouth. Analog data of each image photographed by the photographing camera 1 is A / D converted by the A / D conversion means 3 and then taken into the authentication means 4 and taken into the biometric identification means 5.

  The authentication unit 4 has an image processing function for extracting a face part image (face image) from the upper body image of the photographed person H, and template matching of the extracted face image with respect to the template image registered in the holding unit 1. For example, using a face image extracted from at least one frame image captured first, and performing template matching on the template image registered and held in the holding means 6, and the person of the face image And a function for determining whether or not the person corresponds to a registered template image. If the determination result, the face image used for face authentication, and the registrant are determined, the template image is displayed. Correspondingly registered personal information is output to the determination result output means 7 as personal information for identifying the photographed person H.

On the other hand, the biometric recognition unit 5 extracts each face image from two consecutive frames of images, and further creates an image difference between the frames of the extracted face image. when the count value by counting the number of changed pixels in the estimated region with each region has exceeded a preset threshold value ran eye if the eye region and (there is blinking), also In the mouth area, it is determined that the mouth has moved, and if only one of the movements is determined, the subject is recognized as a living body, and both movements or both do not move together. Has a function of recognizing that the subject is not a living body, and outputs the recognition result to the determination result output means 7.

  In other words, when impersonation is attempted using a face photo or mannequin, the mouth and eyes move simultaneously when they are moved. Therefore, the inter-frame difference image created as described above is an image indicating that they have moved at the same time, or both of them are still images. Therefore, the biometric recognition means 5 is not easily a living body from these inter-frame difference images. Can be recognized.

  Further, the determination result output means 7 holds as evidence of the determination result when the face of the subject is identified as a living body and the face of the person who desires to enter the room is authenticated as a pre-registered person's face. The means 6 performs processing for storing an input image, information for identifying an individual, and a biometric identification result. Similarly, even when it is identified that the subject is not a living body, if the face of the subject matches the face of a person registered in advance by face authentication, the person is specified, and the biometric identification result and the input image are identified. Then, a process for storing in the holding means 6 information identifying an individual who is the face of the subject is performed. With this stored data, the administrator can later confirm the person who was impersonated, and the threshold for biometric identification determination and authentication determination can be changed based on the past impersonation history, so it is highly reliable Security can be achieved.

  In addition, when an identification result indicating that the subject is not a living body is obtained, a plurality of imaging cameras installed in the vicinity of the face authentication device can be used to capture an image of a person holding a photograph or the like near the face authentication device as an evidence image, or face authentication The face authentication device can be operated by providing a wall that cuts off the exit of a person holding a photograph or the like near the device.

  Here, FIG. 2A and FIG. 2C show the input image before blinking and the upper body image when blinking, and FIG. 2B and FIG. A face image from which a face portion is extracted is shown by a frame (shown by a broken line). In this case, an image of a difference between frames of both face images is as shown in FIG.

  3 (a) and 3 (c) show an image of the upper body before moving the mouth and an image of the upper body when the mouth is moved, and FIGS. 3 (b) and 3 (d) show the images from the respective images. A face image obtained by extracting a face portion with a square frame is shown, and an image of a difference between frames of both face images in this case is as shown in FIG.

  4 (a) and 4 (c) show two images of the upper body before and after the mouth and eyes are not moved, and FIG. 4 (b) and FIG. 4 (d) show square images from the respective images. A face image obtained by extracting a face portion with a frame is shown, and an image of the difference between frames of both face images in this case is as shown in FIG.

  Here, the image used for the inter-frame difference is an image obtained by binarizing a grayscale image. The white pixel of the difference image indicates a pixel having a difference value of “1”, and the black pixel indicates a pixel having a difference value of 0 or negative. Therefore, the white pixels of the mouth and the eye of the difference image are counted and compared with a preset threshold value. The face image may be a grayscale image or a differential image.

  Note that a well-known technique may be used as the face image extraction method, and detailed description thereof is omitted here. For example, an average face image (see FIG. 5) determined in advance can be used from the imaging camera 2. There are a method of performing template matching to the entire input image, a method of performing template matching by enlarging / reducing or rotating the template image, and a method by Hough transform in order to cope with rotation and size change of the face.

  The eye position / mouth position for counting the number of pixels is, for example, based on a predetermined average face image (see the eye position and mouth position in FIG. 6A) (see FIG. 6B). An appropriate method may be used such as a method in which the face image is extracted from the input image shown in c) in consideration of the rotation angle and the magnification, and is not particularly limited, as shown in FIG. FIG. 6E shows an example corresponding to the movement of the mouth and the movement of the mouth.

  Further, in the above-described case, the face image is used as the image for performing the inter-frame difference. However, the inter-frame difference in FIG. 7C is used by using the images (FIGS. 7A and 7B) captured by the imaging camera 2 as they are. Even when an image is created, the above-described eye movement (blink) and opening / closing operation of the mouth can be detected. In the inter-frame difference image of FIG. 7C, it is detected that the mouth is moving.

In this embodiment, the eyes and mouth are used as a representative example of the part that can be easily changed in shape in the face. However, if the combination of two parts that are easy to move on the face is used, For example, a combination of a mouth and eyebrows (lifting eyebrows, bringing eyebrows close) may be used.
(Embodiment 2)
In Embodiment 1 described above, the presence / absence of movement of the mouth and eyes is detected from the inter-frame difference image to make a living body determination. However, most of the photographs used for impersonation are printed or printed on rectangular photographic paper or paper. Therefore, the biometric identification means 5 of the present embodiment is characterized in that it has a straight line detection function for determining that it is not a photograph, that is, a living body, by detecting a rectangular frame based on the interframe difference. In this case, the difference between frames may be a grayscale image obtained by imaging with the imaging camera 2 or may be an image after differentiation processing, and may be a binary image or a multivalued image. Further, when the function based on the Hough transform is adopted as the straight line detection function, the Hough transform is performed on the binary image. Therefore, if the inter-frame difference result image is a multi-value image, a binarization process is performed. The Hough transform is a straight line detection method based on general image processing, and the presence / absence of a straight line and the length of a straight line portion are determined based on the set degree of voting results in the ρ-θ space in the Hough transform. 8A and 8B are images used for inter-frame differences, and FIG. 8C shows an inter-frame difference image.

  Thus, if the biometric identification unit 5 detects a straight line from the inter-frame difference image, it determines that the face image is a photograph, that is, determines that the face image is not a living body.

  According to the present embodiment described above, since the detection of a straight line that does not exist near the face of a living person, that is, the detection of a frame of a photograph, the living body identification can be correctly determined.

Since the circuit configuration of this embodiment is the same as that of the first embodiment, illustration and description thereof are omitted.
(Embodiment 3)
In the first embodiment, the presence or absence of eye or mouth movement is detected from the inter-frame difference image to make a living body determination. In the second embodiment, the living body is determined based on the presence or absence of a straight line. The biometric identification means 5 is characterized in that a method of determining that a moving object is present near the extraction position is not a living body when it is detected from the inter-frame difference image. I to this Riutsushi true is valid if it is not a straight line or a circle, or, it is possible to detect the hand holding the photo.

Since other circuit configurations of the present embodiment are the same as those of the first embodiment, illustration and description thereof are omitted.
(Embodiment 4)
The present invention is designed to perform face authentication of a person shown in a photograph even when impersonation is performed by a photograph, but in the case of a photograph, ambient light is easily reflected by the gloss of the paper surface. since, darker brighter place, or, depending on the contrary, it may be impossible to obtain a clear image by the imaging camera 2 if there is the shading relationship varies greatly. In addition, for example, when printing or printing on A4 size photographic paper or paper, it is conceivable that the face portion captured by the imaging camera 2 is blurred and unclear.

  When face authentication is performed using such a blurred image, there is a high possibility that the correlation value decreases and it is difficult to identify an individual.

  Therefore, in the present embodiment, when an identification result indicating that a living body is not a living body is output by the living body identifying means 5, a template matching threshold value for performing an authentication determination as to whether or not a person is registered in advance is used when the body is not a living body such as a photograph. This is different from the first embodiment in that an authentication threshold value setting means 9 for setting a threshold value to be lowered is provided as shown in FIG. Since other circuit configurations of the present embodiment are the same as those of the first embodiment, illustration and description thereof are omitted.

Thus, in the present embodiment, when an identification result indicating that the subject is not a living body is obtained, the person in the photograph can be specified by face authentication by lowering the face authentication threshold in the authentication unit 4 by the authentication threshold setting unit 9. It can be done reliably. In addition, since face authentication is performed separately from biometric identification, entry to the room is not permitted even if face authentication is performed with a lower threshold.
(Embodiment 5)
In the present embodiment, it is assumed that a person who has tried to make the face authentication apparatus make an erroneous determination using something other than a living body such as a photograph tries to enter the room again, as shown in FIG. It is characterized in that an identification threshold setting means 10 for setting the identification threshold of the means 5 is provided. Since other circuit configurations of the present embodiment are the same as those of the first embodiment, illustration and description thereof are omitted.

  Thus, based on the face authentication result of the person H who is about to enter this time, the authentication means 4 receives the personal information of the person H, and the determination result output means 7 impersonates the person using a photograph or the like in the past. Whether or not the person is a person is determined from the stored data of the holding means 6, and when the determination result is the person, the identification threshold value setting means 10 is controlled so as to increase the threshold value for identification of the biometric identification means 5. To do.

Here, when it is determined that the person has been impersonated using a photograph or the like in the past, the identification threshold value of the biometric identification unit 5 is increased, so that biometric identification is strictly performed. Even if impersonation using is performed, the possibility that the person is mistakenly recognized as a living body and permission to enter the room is reduced, and security can be maintained.
(Embodiment 6)
This embodiment assumes the case where a person who tried to make the face authentication device make a wrong determination using something other than a living body such as a photograph as in Embodiment 5 tries to enter the room again, The determination result output means 7 that has received the personal information of the person from the authentication means 4 based on the face authentication result of the person who wants to enter the room uses the information in the holding means 6 to impersonate the person in the past using a photograph or the like. If it is determined that the person is a person who has been lit, the threshold value of the authentication means 4 under the control of the determination result output means 7 is determined to be strict. The authentication threshold value setting means 9 ′ is set as shown in FIG. 11 so as to make it easier to determine that the person is not a person even if the result of the person identification is a living body. This is different from the fifth embodiment. Since other circuit configurations of the present embodiment are the same as those of the first embodiment, illustration and description thereof are omitted.

  Thus, in the present embodiment, when it is determined that the person who is to be face-authenticated this time is a person who has been spoofed using a photograph or the like in the past, the biometric identification means 5 is identified as a biometric. Even if the result is output, the determination result output means 7 does not immediately output the determination result permitting entry, but raises the face authentication threshold of the authentication means 4 by the authentication threshold setting means 9 ′, and the authentication means 4 again Face authentication is performed, and when it is authenticated as a registered person in this second face authentication, a determination result for permitting entry is output.

  In other words, by tightening face authentication for a person who has been impersonated in the past, the possibility of giving an incorrect entry permission to a malicious person is eliminated, and security is maintained.

  In the past, when a person impersonated in the past tries to enter the room, the determination result output means 7 is allowed to enter the room even if it is authenticated as a person registered by face authentication and identified as a living body. The determination result indicating may not be output.

  There was a change in external factors, such as when the person who stored the data re-registers the template image when impersonating the holding means 6 in the past, or when the number of guards was increased in the security of entrance / exit management. In this case, the biometric identification result stored for the person and the past history of information for identifying the individual are deleted. Thereby, the threshold value in the authentication determination biometric identification determination for the corresponding person becomes the same value as that of a normal person.

1 is a circuit configuration diagram of Embodiment 1. FIG. 6 is an explanatory diagram of biometric authentication based on inter-frame differences using face images in Embodiment 1. FIG. 6 is an explanatory diagram of biometric authentication based on inter-frame differences using face images in Embodiment 1. FIG. It is explanatory drawing of the biometric authentication by the difference between frames using the face image in Embodiment 1. FIG. FIG. 3 is an example of an average face image used for face image extraction in the first embodiment. FIG. 5 is an explanatory diagram for setting the eye and mouth area for counting the number of pixels at the time of biometric authentication based on a difference between flakes using a face image in the first embodiment. It is explanatory drawing of the biometric authentication by the difference between frames using the input image in Embodiment 1. FIG. It is explanatory drawing of the identification method of the face photograph at the time of the biometrics authentication by the difference between frames in Embodiment 2. FIG. 6 is a circuit configuration diagram of a fourth embodiment. FIG. 9 is a circuit configuration diagram of a fifth embodiment. FIG. 10 is a circuit configuration diagram of a sixth embodiment.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Holding means 2 Imaging camera 3 A / D conversion means 4 Authentication means 5 Biometric identification means 6 Holding means 7 Determination result output means 8 Voice generation means H Person

Claims (6)

Template matching between an imaging camera that continuously captures images including the face of the person to be authenticated for a plurality of frames, an image of at least one frame from the imaging camera, and an image of a person registered in advance as a template image The authentication means for determining whether or not the person to be authenticated is registered, and the presence or absence of movement of a plurality of parts on the face from the images of the plurality of frames photographed by the photographing camera to detect the subject of the photographing camera A person to be authenticated is registered only when a biometric identification means for identifying whether or not a living body is present and an authentication result indicating that the authentication means is a registered person and an identification result that the biometric identification means is a living body are output A determination result output means for outputting a determination result indicating that the person is a person, and the recognition means regardless of the identification result of the biometric identification means. And a holding means for storing information for identifying the registered person when authentication target person is authenticated as the registered person, the biometric identification means, a face from the image of a plurality of frames wherein the imaging camera has photographed The position is extracted, and the number of change pixels in the eye area and the mouth area is counted from the inter-frame difference image of the face position image, and the count number exceeds a preset threshold corresponding to each area. A face authentication apparatus that outputs a discrimination result indicating that it is a living body when it is determined that there is movement when it does not exceed, and there is no movement when only one of both areas is moving . The biometric identification unit has a straight line detection function for detecting a straight line from an inter-frame difference image created from a plurality of frames taken by the photographing camera, and the straight line detection function detects a straight line having a predetermined length or more. 2. The face authentication apparatus according to claim 1, wherein an identification result indicating that it is not a living body is output. The biometric identification unit extracts a face position from each image of a plurality of frames taken by the imaging camera and creates an inter-frame difference image, and the inter-frame difference image has a predetermined number of change pixels near the outside of the face position. The face authentication apparatus according to claim 1 or 2 , wherein if it is above, an identification result indicating that it is not a living body is output. 4. The face according to claim 1 , further comprising an authentication threshold setting unit configured to perform authentication by lowering an authentication threshold of the authentication unit when the identification result of the biometric identification unit is not a living body. 5. Authentication device. In no event identification result of the biometric identification device is not a living body, person the authentication unit is authenticated as a person being registered, when it becomes again the authentication target person, the biometric identification means for the person 5. The face authentication apparatus according to claim 1, further comprising an identification threshold setting unit that increases the identification threshold . When the identification result of the biometric identification unit is not a biometric and the person who has been authenticated as the person registered with the authentication unit becomes a person to be authenticated again , the face authentication threshold of the authentication unit is increased. face authentication equipment according to any one of claims 1 to 5, characterized in that an authentication threshold value setting means to perform authentication Te.